Available online at www.sciencedirect.

com

ScienceDirect
Transportation Research Procedia 18 (2016) 297 – 304

XII Conference on Transport Engineering, CIT 2016, 7-9 June 2016, Valencia, Spain

Managing airlines: the cost of complexity

César Trapote-Barreira a *, Andreas Deutschmann b, Francesc Robusté a
a Technical University of Catalonia – BarcelonaTech, Barcelona, Spain
b Deutsches Zentrum für Luft- und Raumfahrt, Braunschweig, Germany

Abstract

This paper is dedicated to the structure of airline networks as a sink of efficient airline operations. Parameters of complexity were
derived and mirrored on level of service as well as efficiency parameters. Airlines usually considerers an operational overhead to
predict the total flight operation cost. This parameter includes the expected cost for disruptions and delays. When an airline has to
mobilize an aircraft in a base for recovering the service or for breaking an emergent dynamic, then it is running extra costs. The
cost of managing complexity in the airline industry has a direct impact on profit and loss account. Therefore, this paper presents
an integrated approach to evaluate this cost, based on padding and aircrafts dedicated to recover disruptions. Finally, some
additional indicators are derived to evaluate reliability improvement as part of complex performance.
© 2016 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Elsevier B.V. All rights reserved.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-reviewunder
Peer-review under responsibility
responsibility of organizing
of the the organizing committee
committee of CIT of CIT 2016.
2016
Keywords: Airline, network, cost, complexity, reliability, delay, propagation, performance.

1. Introduction

Complex systems have always existed, but complexity has gone from something found mainly in large systems,
such as cities, to something that affects a lot of common things and organizations. Systems that used to be separated
are now interconnected and interdependent, which means more complexity in many cases. Furthermore, technology
revolution has been a factor to increase level of complexity.
Complex organizations are far more difficult to manage than merely complicated ones. It is harder to predict what
will happen, because complex systems interact in unexpected ways. Also, it is harder to make sense of things, because

* Corresponding author. Tel.: +34 934 011 947.

E-mail address: cesar.trapote@upc.edu

2352-1465 © 2016 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of CIT 2016
doi:10.1016/j.trpro.2016.12.039
298 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304

the degree of complexity may lie beyond cognitive limits. And it is harder to place bets, because the past behavior of
a complex system may not predict its future behavior. In a complex system the outlier is often more relevant than the
average.
Three properties determine the complexity of an environment. The first, multiplicity, refers to the number of
potentially interacting elements. The second, interdependence, relates to how connected those elements are. The third,
diversity, has to do with the degree of their heterogeneity. It is possible to understand both simple and complicated
systems by identifying and modeling the relations between the components of the network; the relations can be
reduced to clear and predictable interactions. It is not possible to understand complex systems in this way, because all
the elements are interacting continuously and unpredictably.
This paper aims to develop an analysis of the airline network from the point of view of complexity science. It is
structured in three sections. The first includes an introduction to complex theory and its applications to airline industry.
The second section addresses a network analysis by applying complexity indices is presented. Finally, a methodology
of reliability estimation and cost associated to delay propagation is developed.

2. State of the art

The analysis of the spatial configuration of networks was put in the center of empirical investigation in the field of
operations research and quantitative social science research for many years. Much attention has recently been paid to
the study of network properties emerging in many social, spatial and economic fields, as witnessed by the vast amount
of literature published in the past years (Barthélemy, 2003; Reggiani et al., 2006).
Air transport shows indeed clear network features, which impact on the way single airline carriers operate (Button
et al., 2000). The abundant scientific literature on airline networks has addressed this topic in terms of theoretical
modeling and empirical measurements on different typologies of airline network configurations. This strand of recent
research aimed to measure the network structure in relation to the new trends in airline business strategies denoted as
‘low cost’ principles.
In this context, research has emerged that mainly addressed the issue of describing and classifying networks by
means of geographical concentration indices of traffic or flight frequency (Caves et al., 1984; Reynolds-Feighan,
2001). These measures, such as the Gini concentration index (Gini, 1912), provide a proper measure of traffic
concentration of the main airports in a simple and well-organized network. However, if a real-world network structure
is complex the concentration indices may record high values for all types of structure, but fail to clearly discriminate
between different network shapes (Alderighi et al., 2007). There is a need for a more appropriate measurement of
connectivity structures in complex networks.
However, indices derivated from spatial network analysis are static and often a dynamic perspective of network is
necessary. One of the main problems in the field of airline management is delay propagation. Delays have a strong
impact on operational reliability and these impacts directly on profit and loss account and passenger experience.
Delay propagation is the result of different factors, including the lack of coordination of airline flight schedules,
finely tuned airline flight schedules with little slack to dampen delay propagation, high levels of congestion preventing
re-accommodation of delayed flights, or high aircraft load factors preventing timely re-accommodation of passengers
who misconnect or whose flights are cancelled. All delay causes combined create passenger disruptions and lengthy
passenger waiting times that create new flight delays.
According to Eurocontrol (2014) there were 1.7% more flights per day in the reference area than in 2013. However,
data received directly from airlines by CODA describing delays from all-causes illustrated a stable situation for the
network during the year. First, the average delay per delayed (ADD) flight of 26 minutes per flight; this was a small
decrease of 2.6% compared to 2013. Second, this small improvement was offset by a small increase of 1.3% to 37.4%
of flights delayed on departure (>=5 minutes). Third, the share of reactionary delay was 44% of delay minutes reported
compared to 45% in 2013. Finally, regarding arrival delay, the average delay per delayed flight on arrival from all-
causes was 27.2 minutes per flight in 2014. Then, the percentage of delayed flights increased by 0.7% to 34.3% and
operational cancellations remained stable at 1.5% of planned flights.
Delays cause immense losses to the Air Traffic System, a situation that will worsen in the near future if traffic
increases. Models allowing stakeholders to characterize mechanisms behind delay propagation, to forecast network
congestion, and to optimize planning and operational practices to mitigate delays are thus of great relevance.
 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304 299

Researchers who have been studying the performance of ATM have done a significant effort to identify the causes
for initial or primary delays. These primary delays can in turn trigger a cascade of secondary delays by the introduction
of a ripple effect (Jetzki, 2009). Because of the inherent complexity of the mechanisms that produce and boost delay
spreading, different modeling techniques were proposed. A first line of research focused on simulating the air traffic
system as a network of queues without considering information on aircraft schedules (Schaefer and Miller, 2001). A
second line of research was devoted to analytical approximations for modeling the airport as a dynamic queuing
system (Pyrgiotis et al., 2013). Additionally, statistics can be used to predict delay patterns (Churchill et al., 2007; Xu
et al., 2005).
Deutschmann (2012) proposes a method to predict off-block delays based on assumptions of non-linear physics
and knowledge of airport dynamic characteristics, applying this method to several US airports. Later, the author
(Deutschmann, 2013) carried out an analysis and development of models to understand the airport as a dynamic
system, applying the concept of harmonic oscillating system. Trapote-Barreira (2015) proposes a complete analysis
of airline network configuration, integrating static perspective through a set of complex indices and dynamic
perspective with a delay propagation algorithm. This work is the foundation of this paper.
Technologically driven transport systems are characterized by a networked structure connecting operation centers
and by dynamics ruled by pre-established schedules. Schedules impose serious constraints on the timing of the
operations; they condition the allocation of resources and define a baseline to assess system performance. Technical,
operational or meteorological issues affecting some flights give rise to primary delays. When operations continue,
such delays can propagate, magnify and eventually involve a significant part of the network. Metrics have been defined
to quantify the level of network congestion. The results indicate that there is a non-negligible risk of systemic
instability even under normal operating conditions. Passenger and crew connectivity were also identified as the most
relevant internal factor contributing to delay spreading.

3. Analysis of network complexity

This section aims to investigate the scientific potential and applicability of a series of network connectivity and
concentration indices, in order to properly typify and map out complex airline network configurations.
Modeling complex networks is also a great challenge: on one side, the topology of the network is governing the
complex connectivity dynamics; on the other side, the functional-economic relationships in such networks might also
depend on the type of connectivity structure. The understanding of these two interlinked network aspects may be
instrumental for capturing and analyzing airline network patterns. Then, the focus of this work is one airline and how
indices can provide information for decision-making processes.
The analysis of complex systems as complex networks has been a growing trend in the last years (Strogatz, 2001).
However, one of the most relevant contributions was done by Gini (1912). Moreover, complex networks have been
crucial in order to understand many emergent phenomena in systems with a large number of interacting actors. Zanin
and Lillo (2013) have successfully applied this formalism to the study of air transport.
The paper presents an empirical and short case study, applying indices of complexity to real data. Due to an
agreement of confidentiality the identity of these airlines are not revealed. Then, Airline 1 is a short-medium haul low
cost carrier; and it is a small and start-up project. This airline manages a point-to-point network, but very concentrated
in three bases around Europe. It is operating more than 300 flights in one week and more than 60 airports. Airline 2
is a large low cost carrier, focused on short and medium haul, and one of the most important LCC in Europe. It
manages more than 180 airports (more than 70 bases) and more than 2,000 relations each day.

3.1. Indices of complexity

Airline networks may exhibit simple or complex topologies. Networks have been given several definitions in the
framework of graph theory. In this context it is useful to outline some indicators most frequently used to represent the
network shape (Barrat et al., 2004). Table 1 synthetizes a set of indices of complexity.
Figure 1 shows the degree index and strength index for airline 1 and Figure 2 shows them for airline 2. The level
of concentration is higher than concentration of airline 1 and it is due to the fact that airline 2 dominates some markets
and it supplies high frequencies in some routes very profitable. It is not surprising that both low cost airlines show
300 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304

distributions relatively concentrated despite of the fact they manage routes point to point. First of all, from a business
point of view, airlines have the need to concentrate main operations in a few bases because it is easier to control the
cost. Furthermore, it is easier to give reliability to the network. Second, one of the critical aspects it is to not provide
connections for passengers, this issue allows managers to not transfer delays in connection times. In practice, low cost
airlines manage the best indices of punctuality.
Cluster coefficients are more interesting for this analysis and comparatively; airline 1 presents a distribution with
fewer tendencies to clusterization. The main reason is that this airline manages only a few bases around Europe and
less flights than airline 2. With density and radial configurations, these networks tend to create clusters. Network
degree distribution and average nearest-neighbors degree for airline 1 and 2 are presented in Trapote-Barreira (2015).
Summarizing, network degree distribution shows an asymmetric level of supply and spatial distribution, especially
for large airlines (this fact can be related with the schedule design), and average nearest-neighbors degree allows
recognized patterns of concentration much more important for airline 2 than airline 1.
Gini index has been calculated for both airlines, being G1=0.6310 and G2=0.6422. Taking into consideration that
they are low cost airlines with point-to-point configurations, the level of concentration is very important.

Table 1. Indices of complexity.

Indicator Equation Definition
Measure of centrality, level of connectivity of each node
Degree Index
݇௜ ൌ ෍ ܽ௜௝  in terms of links (adjacency matrix).
(Barrat et al., 2004) ௝
Matrix of incidence, A={aij} for i, j = 1…n.
Strength Index Measures connectivity of each node in terms of flights
‫ݏ‬௜ ൌ ෍ ‫ݓ‬௜௝ ܽ௜௝ 
(Barrat et al., 2004) ௝ (weight wij for relation aij).
Measures connectivity in terms of probability of finding
Vertex Degree Index ܰሺ݇ሻ nodes with k links.
ܲሺ݇ሻ ൌ 
(Candarelli, 2007) ݊ N(k) is the number of airports with k connections.
Cluster Coefficient ‫ܧ‬௜ Gives information about the spatial structure and depends
‫ܥ‬௜ ൌ  on number of triangles in the network.
(Kaiser, 2008) ݇௜ ሺ݇௜ െ ͳሻ
Average Nearest-neighbors Gives information about correlation of the degrees of
Degree ͳ
݇௡௡ǡ௜ ൌ ෍ ݇௝  connected nodes (nodes with high degree tend to connect
݇௜ ௝ with high degree nodes).
(Barrat et al., 2004)
Weighted Average
Nearest-neighbors Degree ௪
ͳ
݇௡௡ǡ௜ ൌ ෍ ‫ܣ‬௜௝ ‫ݓ‬௜௝ ݇௝  Idem, with weights related to flights.
‫ݏ‬௜ ௝
(Barrat et al., 2004)
Gini Index σ௜ σ௝ ห݂௜ െ ݂௝ ห
‫ܩ‬ൌ  Measures geographical concentration of flights
(Gini, 1912; Cento, 2006) ʹ݊ σ௜ ݂௜

Fig. 1. (a) degree index; (b) strength index for Airline 1.

 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304 301

Fig. 2. (a) degree index; (b) strength index for Airline 2.

Fig. 3. (a) cluster index for Airline 1; (b) cluster index for Airline 2.

4. Evaluation of operating cost related to improved reliability

Indices of complexity allow understanding the network at strategic level. However, they are a static perspective
that cannot be applied to understand the dynamic performance of the operation. For this reason, this section presents
an algorithm to control airline planning and to manage delay propagation, taking into consideration the cost of
reliability in the operational field.

4.1. Problem statement

Consider a flight scheduling and aircraft routing. This is the basis for a propagation tree scheme. Delays, in this
model, follow statistical distributions. First, flight time can present an average delay μF, given in minutes per one hour
of flight, and a statistical deviation σF, considering N(μF, σF )=N(1.78, 3.30), given in minutes per one hour of flight.
Second, turnaround time can be delayed in the same way, with average delay μG per one hour of ground time and σG
statistical deviation, N(μG, σG )=N(2.32, 3.43) in minutes per hour. This basic assumption improves simplicity and it
is consistent from a conceptual point of view.
Airlines can plan or test their flight schedules with characteristic flight and turnaround times, where k is the
parameter of reliability that they want to considerer to hedge their operation (part of the deviation to absorb).
Moreover, airports and airlines have the possibility to reduce a certain amount of delays due to some operational
actors. However, this work does not consider this aspect, so the propagation is considered inelastic through the same
route. Furthermore, if the airport is congested, it can transfer delays. This transferability is modeled with a parameter
that depends on level of utilization of the airport and hubbing practices of the airline. Airline network configuration
is a key aspect to understand delays propagation.
A flight scheduling is considered (FS), which is planning taking the flight time between two scheduled airports and
the turnaround time into account. When delays appear, this FS is degenerated to FS’, which is the timetable after the
perturbation. Airline can design a flight schedule FS* with padding to be reliable in front delay propagation. FS* can
be the same flight schedule with departure and arrival time adjusted to new conditions or it can be designed adhoc.
302 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304

This work considers the adjusted version that consists on introducing extra time for flight time and turnaround time
to absorb mean values of delays and one or more times the standard deviation. Then, the parameter K decides the level
of coverage that airline wants to have and it results in different values of characteristics delays. It is not allowed any
value of K because fleet is constrained. Also, a set of routes ܴ ൌ ൛‫ݎ‬௝ ൟ are considered and it is related to first assignment
done for a flight schedule (10 flights), where each route ‫ݎ‬௝ containts a sequence of flights that are served by the same
plane.

4.2. Cost of operation considering delays

First, operating costs increase with delays and padding, but in a different way. For unplanned delays, the cost for
airline is evaluated in terms of variable costs (because fixed costs not vary in this scenario) and them can be estimated
with expressions proposed by Trapote-Barreira (2015). However, a coefficient of penalization is considered because
the perturbation enforces to reassign resources and extend duty time for crew. Second, passengers are penalized in
their travel time because extra time not expected is necessary. Then, passenger costs are estimated as proportion of
value of time related to this increment of time. If padding is considered, then more resources are going to be needed
during planning phase (or not) but these costs are going to be balanced by savings on penalization. Furthermore,
passengers are going to be beneficiated with more reliability.

4.3. Backup of resources for reliability

Airlines usually take the need into consideration of having backup resources in main bases to recover the planned
service or mitigate perturbations in flight scheduling. When flight schedule is delayed and these delays exceed a
threshold, airline can make the decision of breaking a route into two sub-routes and the second one departs on time
with the backup. In this kind of operation, eventually, a ferry flight can be necessary to send it to the airport where
this is necessary. These kinds of flights are not desirable by companies because of the expected costs. However, if
penalization increases, then this strategy can be interesting to solve operational problems. Furthermore, costs of delays
are variable costs, but backup resources are fixed and variable costs. Therefore, this extra cost (and padding strategy)
decreases margins in profit and loss account. These are costs of reliability and it can make the difference between
having negative or positive balance of accounts at the end of the year.

4.4. Algorithm of control

Assume a given flight schedule and quantification of time for flights and turnarounds. An algorithm of control is
in charge to monitories the development of operation. Considering a delayed flight is, the following flight that is
operated by the same plane it is very likely that is flight is being delayed due to low buffer times and transferability
of delays. Except that padding strategy can absorb this delay or additional buffer is incorporated. Furthermore, in case
of hubs, transferability of delays between different airplanes is possible because there is the coordination of
connections (transfer passengers -usually LCC do not have connections-).
The algorithm considers each flight of FS. However, it starts computing all delays according with distributions
(section 4.1) and it applies a mechanism of propagation (described above) and control. This mechanism opens a
channel to operators to introduce actions to recover the plan or mitigate delays. For each arrival, the algorithm
evaluates the prevision of delays and it can decide:
1. To mobilize a backup. If predicted delay in a route exceeds a threshold (τ), one backup is mobilized,
recovering the original SDT for the following flight. Only if airline has this resource in the base, else the
airline can send it where is necessary with an extra cost.
2. To reorganize routes. If another route accumulates less delays and the swap is allowed (coincidence of flights
at the same airport at the same time). Changing the routes may increase the cost due to reallocating crew and
flights at the end of the cycle with ferry flights or recover the original end of routes if it is possible.
3. To permit the delay propagation. If other alternatives are not possible, then it is necessary to accept the delay
propagation. This alternative would give the option to work at level of flight plan and turnaround operations to
absorb a certain amount of the delay (which is not the goal of this work).
 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304 303

The algorithm runs a simulation based on probability distributions. The total cost of operation with compensations
(or penalizations) is evaluated and it defines the criteria to make the final decision. Finally, for each pair of variables
K and number of aircrafts in the backup, the operator finds a trade-off to improve reliability of FS.

4.5. Numerical experimentation

Given a flight scheduling FS0 and distributions for flight delays N(μF, σF )=N(1.78, 3.30) and turnaround delays
N(μG,σG )=N(2.32, 3.43), both in minutes per hour. The airline decides the padding and estimates FS R for different
scenarios: 1. no delays considered, 2. delays with K=0, 3. K=1, 4. K=2 and 5. K=3., which produces different routing
strategies (Trapote-Barreira, 2015). A preliminary evaluation of cost of delays is carried out without applying the
algorithm of delay propagation (without mechanisms of active control, Table 2). The costs are related to the level of
the network reliability.

Table 2. Results of scenarios for delay propagation without active control.

Scenario 2. Scenario 3. Scenario 4. Scenario 5.
Scenario 1.
Delays Delays Delays Delays
Indicator No delays
considered considered considered considered
considered
with k=0. with k=1. with k=2. with k=3.
Fleet 3 3 3 3 3
5 5 5 5
Original cost planned (EUR) 1.3689·10 1.4066·10 1.4754·10 1.5443·10 1.6131·105
Accumulated delay for airline (h) 1.09 0.38 0 0 0
3 3
Extra cost of delays for airline (EUR) 4.9942·10 1.7347·10 0 0 0
Accumulated delay for passengers (h) 605.69 300.44 36.49 6.79 0
Cost of delays for passengers (EUR) 3.6342·104 1.8026·104 2.1897·103 407.6793 0

Table 3. Results of scenarios for delay propagation with active control and threshold of 15min.
Scenario 2. Scenario 3. Scenario 4. Scenario 5.
Scenario 1.
Delays Delays Delays Delays
Indicator No delays
considered considered considered considered
considered
with k=0. with k=1. with k=2. with k=3.
Fleet (backup required) 3 (1) 3 (1) 3 (0) 3 (0) 3 (0)
5 5 5 5
Original cost planned (EUR) 1.3689·10 1.4066·10 1.4754·10 1.5443·10 1.6131·105
Accumulated delay for airline (h) 0.58 0.07 0 0 0
Extra cost of delays for airline (EUR) 2.5815·103 338.86 0 0 0
Accumulated delay for passengers (h) 419.08 191.81 36.49 6.79 0
Cost of delays for passengers (EUR) 2.5145·104 1.1509·104 2.1897·103 407.6793 0
Applying the algorithm of delay control with backup fleet, some disruptions are eliminated or mitigated. Table 3
shows results if different value of threshold time is considered to activate backup resource. The results indicate that
reliability has an extra cost for airline in terms of extra resources in backup to recover the service or mitigate delays
or the cost of padding (allocating more time for each operation). However, this operating strategy has advantages for
both airlines and passengers. Obviously, if an airline has not to pay penalizations, the strategy only generates costs.
Passengers could be the part beneficiated of this strategy because they save a lot of time.

5. Conclusions

The present section summarizes the quintessence of the described research. First, indices of complexity, airline
networks are fascinating examples of emerging complex and interacting structures, which may evolve in a competitive
environment under liberalized market conditions. They may exhibit different configurations, especially if a given
304 César Trapote-Barreira et al. / Transportation Research Procedia 18 (2016) 297 – 304

carrier has developed alliances and has extended its service network. However, this work has analyzed two simple
networks of low cost carriers. One of them is a new entrant in the market and the other is a big carrier.
The presented network is characterized by an important concentration of the activity. The most important reason is
the need to consolidate strong bases and give financial support to airport or airline development plans. This fact is
consistent with the ways of early entrants that manage large networks today. Furthermore, network carriers considered
in the present study exhibit a hierarchical structure despite the service point to point.
Second, related to propagation of delays, the model developed to improve reliability shows that airlines have to
assume higher costs due reactionary delays. These could be very important at the end of each route if there are no
mechanisms to mitigate or eliminate delay propagation. This delay generates costs for airlines and for passengers.
Airlines assume more costs because they have to react against perturbations of flight schedule. However, introducing
padding, backup resources or re-scheduling improves reliability resulting in extra costs. Backup mechanisms obligate
to assume more fixed costs, but increasing the quality of service. Re-scheduling is difficult if there are not coincidences
of planes at the airport. Padding is one option, but costs have to be considered. This strategy runs fine for point-to-
point networks if ferry flights are not allowed, while hub-and-spoke could take profit of back-up resources in hubs.

References

Alderighi, M., A. Cento, P. Nijkamp, P. Rietveld, 2007. Assessment of new hub-and-spoke and point-to-point airline network configurations.
Transportation Review, 27, 529–549.
Barrat, A., M. Barthelemy, R. Pastor-Satorras and A. Vespignani, 2004. The architecture of complex weighted networks. Proceedings of the National
Academy of Sciences of the United States of America, 101, 3747–3752.
Barthélemy, M., 2003. Crossover from scale-free to spatial networks. Europhysics Letters, 63, 915–921.
Button, K. and R. Stough, 2000. Air Transport Networks: Theory and Policy Implications. Edward Elgar, Cheltenham. ISBN 1 84064 429 X
Caldarelli, G., 2007. Scale-free networks: complex webs in nature and technology. OUP Catalogue.
Caves, D.W., L.R. Christensen and M.W. Tretheway, 1984. Economics of density versus economies of scale: why trunks and local service airline
costs differ. Rand Journal Economics, 15, 471–489.
Cento, A., 2006. Challenge to the airline industry: emergence of carriers and low cost carriers. PhD Thesis, VU University Amsterdam.
Churchill, A. M., D.J. Lovell and M.O. Ball, 2007. Examining the temporal evolution of propagated delays at individual airports: case studies.
Proceedings of the 7th USA/Europe air traffic management R&D seminar. Barcelona, Spain.
Deutschmann, A., 2012. Prediction of airport delays based on non-linear considerations of airport systems. 28th International Congress of the
Aeronautical Sciences.
Deutschmann, A., 2013. Understanding the dynamic system airport. 13th World Conference on Transport Research Society. Rio de Janeiro, Brasil.
Eurocontrol, 2014. Annual Network Operations Report. Brussels.
Gini, C. ,1912. Variabilità e mutuabilità. Contributo allo studio delle distribuzioni e delle relazioni statistiche. C. Cuppini, Bologna.
Jetzki, M., 2009. The propagation of air transport delays in Europe. Ph.D. Dissertation. Department of Airport and Air Transportation Research
RWTH Aachen University.
Kaiser, M., 2008. Mean clustering coefficients: the role of isolated nodes and leafs on clustering measures for small-world networks. Journal of
physics, 10.
Pyrgiotis, N., K.M. Malone and A. Odoni, 2013. Modelling delay propagation within an airport network. Transportation Research Part C, 27.
Reggiani, A. and P. Nijkamp, 2006. Spatial Dynamics, Networks and Modelling. Edward Elgar, Cheltenham. ISBN 978 1 845542 450 3.
Reynolds-Feighan, A.J., 2001. Traffic distribution in low-cost and full service carrier network in the US air transport market. Journal of Air
Transport Management, 7, 265–275.
Schaefer, L. and D. Miller, 2001. Flights delay propagation analysis with the detailed policy assessment tool. Proceedings of the 2001 IEEE
international conference on Systems and Cybernetics Vol. 2, 1299-1303.
Strogatz, S. H., 2001. Exploring complex networks. Nature, 410, 268–276.
Trapote-Barreira, C., 2015. Methodology for optimal design of efficient air transport networks in a competitive environment. Thesis. Civil
Engineering PhD Program. UPC.
Xu, N., G. Donohue, K. B. Laskey and CH. Chen, 2005. Estimation of delay propagation in the national aviation system using Bayesian networks.
Proceedings of the 6th USA/Europe air traffic management R&D seminar. Baltimore, USA.
Zanin, M. and F. Lillo, 2013. Modelling the air transport with complex networks: A short review. The European Physical Journal, 215, 5–21.